excluding points in logistic regression fit

[Sundaram Sridharan]

I would like to exclude points in a scatter plot while fitting them to a logistic regression fit. I considered tagging them and using the tag column in limit data by expression. But then it removes those excluded points from the plot - I would like for the fit line to change while still being able to see the excluded points in a different shape - is this possible

[Anders Gavare]

Not very pretty, but:

How about using Subsets Put (Subsets) on Shape by, and then check the checkbox to show one logistic regression curve per shape. The drawback is that you will see both curves then, but by enabling labels (at least the "Curve name") you can see which curve is for the tagged subset and which one is for the normal (non-tagged) subset.

If you have a bool column (tag or similar), you could add one custom expression subset called "Tagged to ignore" with that bool column as the expression, and a second custom expression subset which is the logical inverse called "Data to include". Then deselect all of the standard subsets such as Current filtering.

(If you are not using Color in the scatter plot for something else, then you could potentially also put (Subsets) on Color by, and check the One per Color checkbox as well, but then the curve will be the same color as the markers, which maybe is not very easy to see... You may need to experiment.)

[Sundaram Sridharan]

thank you - the second option works for me. if the number of excluded points are minimal then it doesn't draw a second curve. Works to my advantage